by Cory Rauch 2009-03-14 Category: Data_Mining
Do not take this the wrong way, I love the Knicks and I am a life long fan. I recently was fooling around with RapidMiner, a wonderful open source data mining tool. It allows you to string together built in functions and then perform experiments on data. You could for example run a supervised learner on data to create a model for forecasting revenue or doing classification. It also features some plotting views to see results. So I decided to collect some data on my favorite basketball team and run a RuleLearner on the data. My results are below, keep in mind that this dataset is just for this season and does not consider scoring (which obviously affects the outcome). So if you where going to write a program to perform like the New York Knicks this season what would it look like? Below is the conditional statement core to the main loop that runs this team:
if (HI_SCORER = "A. Harrington" and Team_Is_Away = "N")
return "L"; // The team loses everytime with this criteria (0 / 7)
elseif (Team_Is_Away = "Y" and Opp_Winning_Percentage_0.500+ = "Y")
return "L"; // The team has lost 12 games when this is the case (2 / 12)
elseif (Team_Is_Away = "N" and Day_Of_Week = "Wed")
return "W" // Why are we so good on Wednesday? (6 / 1)
elseif (HI_SCORER = "W. Chandler")
return "W" // Interesting (4 / 0)
elseif (Team_Is_Away = "Y" and HI_SCORER = "N. Robinson")
return L // I still like Nate (1 / 3)
elseif (Day_Of_Week = "Sat")
return "W" // Again why are we so good on Saturday? (4 / 0)
elseif (Team_Is_Away = "Y" and Day_Of_Week = "Sun")
return "L" // (0 / 2)
else
return "L" // Yes, the Knicks loose by default (9 / 10)
This logic proved to provide correct results on 48 out of 61 games (~ 80% correct). As a fan, I find it interesting that when Harrington is our high scorer and the team is playing at home that we are winless. Especially when the team has a winning record at home (18-15 at time of writing, 7 of those losses are when Harrington is the leading scorer).
Other ImprovedSource Articles:
JavaScript bracket notation allows you to use strange variable names
Boot Fedora Linux Faster: How I Modified Fedora To Boot In Under 25 Seconds
How to make your Linux desktop look awesome